Vehicle re-identification method based on multi-view matching and local feature fusion
A local feature and re-recognition technology, applied in the field of image recognition, to achieve the effect of improving the accuracy of re-recognition
Active Publication Date: 2021-06-15
BEIHANG UNIV
View PDF7 Cites 4 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
[0007] In view of this, the present invention provides a vehicle re-identification method based on multi-view matching and local feature fusion. The vehicle re-identification method based on multi-view matching proposed in the present invention solves the problem of mutual matching of images from different views, and can effectively Improve the accuracy of vehicle re-identification, and efficiently identify images belonging to the same vehicle from different shooting angles in large-scale spatio-temporal scenes
Method used
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment
[0082] Using cycleGAN, train the generators that generate the rear of the car, and the rear of the car to generate the front of the car, etc., and use these two generators to generate multi-view images from a single perspective of the image in the data set, such as figure 1As shown; Then, analyze the feature reliability of the generated image, define and assign weights to the discriminative features of the original image and the generated image, use the Yolov3 target detection tool to locate local features, and perform feature extraction according to the feature extraction strategy under the generated image Extraction; finally, design a multi-view fusion algorithm, assign weights according to the importance of the view, and obtain feature vectors containing rich view information to achieve accurate re-identification of vehicles. Use Veri776, VehicleID, Vehicle-1M and other databases for vehicle re-identification image datasets.
[0083] The specific process of generating other...
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More PUM
Login to View More Abstract
The invention provides a vehicle re-identification method based on multi-view matching and local feature fusion. The method comprises the steps: training a generator with different view angles mutually converted through employing a generative adversarial network, and generating a multi-view image for a single view angle of an image in a data set through employing the generator; then, carrying out feature reliability analysis on the generated image, and carrying out local salient feature extraction according to discriminative features of the original image and the generated image and a feature extraction strategy under the generated image; designing a multi-view fusion algorithm, and distributing the weights according to the importance degrees of the views, so that feature vectors containing rich view information are obtained, and accurate re-identification of the vehicle is realized. According to the vehicle re-identification method based on multi-view matching and local feature fusion, the problem of mutual matching of different view images of the same target in a cross-space-time scene is solved, and the vehicle re-identification precision is effectively improved.
Description
technical field [0001] The invention relates to the technical field of image recognition, in particular to a vehicle re-identification method based on multi-view matching and local feature fusion. Background technique [0002] Public safety is the basic need of the people. Since video investigation plays an irreplaceable and important role in case detection, public security situation warning, and large-scale group event perception and prevention, the state and society have put forward higher requirements for surveillance video intelligent analysis technology. Vehicle monitoring is one of the cores of video surveillance and the basis for realizing smart transportation and smart security. Relying solely on manpower to identify vehicles under cross-cameras cannot meet the needs of urban traffic management and video surveillance security, and will consume a lot of manpower and financial resources. Therefore, a vehicle re-identification method based on computer vision technolog...
Claims
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More Application Information
Patent Timeline
Login to View More IPC IPC(8): G06K9/62G06N3/04G06N3/08
CPCG06N3/08G06N3/045G06F18/253G06F18/254
Inventor 张永飞吴明杰蔡益武张天宇李波
Owner BEIHANG UNIV